Visual Analytics Tutorial: Machine Learning

PCX is equipped with Machine Learning tools which support and improve the analysis and assess associated risks. 

Select Output Axes: Select output variables by clicking on the axes. The output variable values are predicted using Kriging-based methods.

Select Neutral Axes: Select variables which will be completely excluded from predictions. Click on axes to designate as neutral.

Kriging: Generate new design specifications via Kriging-based interpolation. Choose values for input variables by clicking on the appropriate axes (highlighted in green). Requires selected output axes.

Optimisation: Generate a new datapoint with optimised output value or a product of multiple output variables. The approach is based on a genetic algorithm which uses Kriging to repeatedly generate new design specifications. The user need to specify the number of iterations and the mutation rate as well as whether the output value is supposed to be maximised, minimised, or approach a given value. Requires selected output axes.

Show Variability: Display the distributions of values for each selected dimension as a Gaussian function (popup window).

Show Uncertainty: Display input and output certainty for selected datapoints (popup window). The output uncertainty is propagated via a Kriging-based approach using Chebyshev nodes. All uncertainties are shown as Gaussian distributions. Requires at least one selected datapoint.